Introduction: Metabolomics is increasingly being used in the clinical setting for disease diagnosis, prognosis and risk prediction. Machine learning algorithms are particularly important in the construction of multivariate metabolite prediction. Historically, partial least squares (PLS) regression has been the gold standard for binary classification. Nonlinear machine learning methods such as random forests (RF), kernel support vector machines (SVM) and artificial neural networks (ANN) may be more suited to modelling possible nonlinear metabolite covariance, and thus provide better predictive models.Objectives: We hypothesise that for binary classification using metabolomics data, non-linear machine learning methods will provide superior generalised predictive ability when compared to linear alternatives, in particular when compared with the current gold standard PLS discriminant analysis.Methods: We compared the general predictive performance of eight archetypal machine learning algorithms across ten publicly available clinical metabolomics data sets. The algorithms were implemented in the Python programming language. All code and results have been made publicly available as Jupyter notebooks.Results: There was only marginal improvement in predictive ability for SVM and ANN over PLS across all data sets. RF performance was comparatively poor. The use of out-of-bag bootstrap confidence intervals provided a measure of uncertainty of model prediction such that the quality of metabolomics data was observed to be a bigger influence on generalised performance than model choice.Conclusion: The size of the data set, and choice of performance metric, had a greater influence on generalised predictive performance than the choice of machine learning algorithm.

Background: A lack of transparency and reporting standards in the scientific community has led to increasing and widespread concerns relating to reproduction and integrity of results. As an omics science, which generates vast amounts of data and relies heavily on data science for deriving biological meaning, metabolomics is highly vulnerable to irreproducibility. The metabolomics community has made substantial efforts to align with FAIR data standards by promoting open data formats, data repositories, online spectral libraries, and metabolite databases. Open data analysis platforms also exist; however, they tend to be inflexible and rely on the user to adequately report their methods and results. To enable FAIR data science in metabolomics, methods and results need to be transparently disseminated in a manner that is rapid, reusable, and fully integrated with the published work. To ensure broad use within the community such a framework also needs to be inclusive and intuitive for both computational novices and experts alike.Aim of Review: To encourage metabolomics researchers from all backgrounds to take control of their own data science, mould it to their personal requirements, and enthusiastically share resources through open science.Key Scientific Concepts of Review: This tutorial introduces the concept of interactive web-based computational laboratory notebooks. The reader is guided through a set of experiential tutorials specifically targeted at metabolomics researchers, based around the Jupyter Notebook web application, GitHub data repository, and Binder cloud computing platform.

Introduction: Metabolomics data is commonly modelled multivariately using partial least squares discriminant analysis (PLS-DA). Its success is primarily due to ease of interpretation, through projection to latent structures, and transparent assessment of feature importance using regression coefficients and Variable Importance in Projection scores. In recent years several non-linear machine learning (ML) methods have grown in popularity but with limited uptake essentially due to convoluted optimisation and interpretation. Artificial neural networks (ANNs) are a non-linear projection-based ML method that share a structural equivalence with PLS, and as such should be amenable to equivalent optimisation and interpretation methods.Objectives: We hypothesise that standardised optimisation, visualisation, evaluation and statistical inference techniques commonly used by metabolomics researchers for PLS-DA can be migrated to a non-linear, single hidden layer, ANN.Methods: We compared a standardised optimisation, visualisation, evaluation and statistical inference techniques workflow for PLS with the proposed ANN workflow. Both workflows were implemented in the Python programming language. All code and results have been made publicly available as Jupyter notebooks on GitHub.Results: The migration of the PLS workflow to a non-linear, single hidden layer, ANN was successful. There was a similarity in significant metabolites determined using PLS model coefficients and ANN Connection Weight Approach.Conclusion: We have shown that it is possible to migrate the standardised PLS-DA workflow to simple non-linear ANNs. This result opens the door for more widespread use and to the investigation of transparent interpretation of more complex ANN architectures.

Rationale: Current guidelines do not sufficiently capture the heterogeneous nature of asthma; a more detailed molecular classification is needed. Metabolomics represents a novel and compelling approach to derive asthma endotypes (i.e., subtypes defined by functional and/or pathobiological mechanisms). Objectives: To validate metabolomic-driven endotypes of asthma and explore their underlying biology. Methods: In the Genetics of Asthma in Costa Rica Study (GACRS), untargeted metabolomic profiling, similarity network fusion, and spectral clustering was used to identify metabo-endotypes of asthma, and differences in asthma-relevant phenotypes across these metabo-endotypes were explored. The metabo-endotypes were recapitulated in the Childhood Asthma Management Program (CAMP), and clinical differences were determined. Metabolomic drivers of metabo-endotype membership were investigated by meta-analyzing findings from GACRS and CAMP. Measurements and Main Results: Five metabo-endotypes were identified in GACRS with significant differences in asthma-relevant phenotypes, including prebronchodilator (p-ANOVA = 8.3 × 10⁻⁵) and postbronchodilator (p-ANOVA = 1.8 × 10⁻⁵) FEV₁/FVC. These differences were validated in the recapitulated metabo-endotypes in CAMP. Cholesterol esters, trigylcerides, and fatty acids were among the most important drivers of metabo-endotype membership. The findings suggest dysregulation of pulmonary surfactant homeostasis may play a role in asthma severity. Conclusions: Clinically meaningful endotypes may be derived and validated using metabolomic data. Interrogating the drivers of these metabo-endotypes has the potential to help understand their pathophysiology.